|Student:||Phillipp Leopold Heinz Jende|
|Timeline:||February 2015 - 31 January 2019|
This project aims to improve the position estimation of mobile mapping platforms. Mobile Mapping (MM) is a technique to obtain geo-information on a large scale using sensors mounted on a car or another vehicle. Under normal conditions, accurate positioning is provided by the integration of Global Navigation Satellite Systems (GNSS) and Inertial Navigation Systems (INS).
However, especially in urban areas, where building structures impede a direct line-of-sight to navigation satellites or lead to multipath effects, MM-derived products, such as laser point clouds or images, lack the expected reliability and contain an unknown positioning error.
This project is addressing that problem by employing high-resolution aerial nadir and oblique imagery as reference data to be independent of GNSS measurements in urban areas. To achieve maximum flexibility towards different MM systems, the project is split into two parts. Since not all MM systems employ laser scanners and cameras, and these systems and data demand different approaches, Zille Hussnain is focusing on the utilisation of LiDAR data, whereas Phillipp Jende concentrates on MM images.
This part of the research project focuses on the development of a procedure to estimate the precise and accurate position of a Mobile Mapping imaging sensor. Utilising airborne images as reference data, correspondences between the data sets are sought, which constitute the basis for an adjustment.
Panoramic photographs are acquired every 5 metres along the MM vehicle’s trajectory. Utilising interior and exterior parameters of MM images, a data pre-processing step projects the image onto an arbitrary surface enabling a higher resemblance with the reference images.
Subsequently, low-level features, such as corners or edges, are detected in the MM image. Again, orientation parameters are exploited to enable a confined search of correspondences in the aerial images. Area-based matching techniques will be used to achieve sub-pixel accuracy.
Yielded tie features between ground and aerial data serve as the ground control, whereas correspondences between MM images could correct relative errors along the vehicle’s trajectory (see Figure 1). In a final step, these information are incorporated into a bundle adjustment to correct the MM data.
Figure 1 Overview: Correspondences between ground and aerial data serving as constraints for an orientation update